[lworld] RFR: 8275583: [lworld] C2 fails to scalarize inline types in safepoint debug info in rare cases [v2]

[Tobias Hartmann]

> In rare cases, it can happen that inline types are not scalarized in safepoint debug info and we hit an assert. The problem is that `InlineTypeBaseNode::Ideal` is only called by IGVN if the node (or one of its inputs) has been modified but not if folding of output nodes leads to a direct connection to a SafePointNode. Usually, the node is still scalarized when we visit all inline type nodes via ` Compile::process_inline_types` but there are cases when that's too late and not scalarizing the node early (prevents other nodes from being removed (for example, during macro expansion).
> 
> The fix is to simply move the code to `SafePointNode::Ideal` which will be invoked whenever one of the (debug) inputs changed.
> 
> I also fixed an issue where `FastLockNodes` with an `InlineTypePtrNode` input are not folded and keep allocations alive.
> 
> Best regards,
> Tobias

Tobias Hartmann has updated the pull request incrementally with one additional commit since the last revision:

  Unrelated fix: Use vk type when checking that casts can be pushed through

-------------

Changes:
  - all: https://git.openjdk.java.net/valhalla/pull/568/files
  - new: https://git.openjdk.java.net/valhalla/pull/568/files/d5c4acac..1514473d

Webrevs:
 - full: https://webrevs.openjdk.java.net/?repo=valhalla&pr=568&range=01
 - incr: https://webrevs.openjdk.java.net/?repo=valhalla&pr=568&range=00-01

  Stats: 13 lines in 1 file changed: 9 ins; 3 del; 1 mod
  Patch: https://git.openjdk.java.net/valhalla/pull/568.diff
  Fetch: git fetch https://git.openjdk.java.net/valhalla pull/568/head:pull/568

PR: https://git.openjdk.java.net/valhalla/pull/568